Data Ready

Once you have your SAS data ready for transport, you need to determine a means to deliver it. There are many ways to send data, but you should strive for process simplicity and data security. To keep your data secure and to comply with 21 CFR-Part 11, you need to encrypt your data files for transport. The best encryption you can use is key exchange high-bit encryption software such as PGP, which creates essentially unbreakable files when used properly. Once your data files are encrypted, you can either send them on physical media such as CD-ROM or send them electronically with secure transmission software such as Secure File Transport Protocol (SFTP). If you need to send data to someone once, a CD-ROM is simple enough to produce. However, if you need to send the data repeatedly, then you should use a more automated electronic method of data exchange. Shell scripts and batch files can be written to automate the electronic data transfer process. 

Another scheduling approach is list scheduling, which schedules operations into control steps, one control step at a time. For the current control step, a list of data ready operators is constructed, containing those operators whose inputs are produced in earlier control steps, and that do not violate any resource constraints. This list is then sorted according to some priority function, the highest-priority operator is placed into the current control step, the list is updated, and the process continues until no more operators can be placed into that control step. This process is then repeated on the next control step, until the entire design is scheduled. Two common priority functions are mobility... 

The CSTEP control step scheduler uses list scheduling on a block-by-block basis, with timing constraint evaluation as the priority function. Operations are scheduled into control steps one basic block at a time, with the blocks scheduled in executidepth-first traversal of the control flow graph. For each basic block, data ready operator are considered for placement into the current control step, using a priority function that reflects whether or not that placement will violate timing constraints. Resource limits may be applied to limit the number of operators of a particular type in any one control step. 

The Sehalloc algorithm maintains two sets of operations a ready-list, which is the list of operations that are data ready, and a max-micro-group, which is a set of operations that can be executed in one control step. Initially, the max-micro-group is empty, and the ready-list contains all operations that are data ready for the first control step. A branch-and-bound, recursive algorithm is then used for scheduling at each step in the algorithm, all possible max-micro-group and ready-list pairs are determined, Splicer is used to perform the data path synthesis for each, the best solution is retained, and the process repeats until the ready-list is exhausted or the cost exceeds the cost of the best solution. 

In basic list scheduling as is done for microcode compaction, an operator cannot be scheduled in a control step until all of the operators that provide it with inputs are scheduled in previous control steps. An operator that meets these conditions is said to be data ready. To support cascaded operator scheduling, this definition must be modified. An operator can be placed in the same control step as an operator that provides it with an input if the accumulated delay does not exceed the clock period. With these modifications, an operator Xa in basic block Rj is data ready for placement into control step csij when for each input ia,b a one of the following conditions is met ... 

Search oplist for data ready operators, remove... 

Scheduling proceeds iteratively until all three operator lists are empty. The outer loop of the algorithm creates new control steps in increasing order and continues the scheduling of multiple step operators that were started in preceding steps. The inner loop of the algorithm selects operators from the data ready list drlist and schedules them into each control step until no further operators can be placed in that control step. As operators are scheduled, new operators become data ready and are removed from oplist and placed in drlist. 

The extended definition of the data ready condition supports the scheduling of cascaded operators by allowing cascaded operators to be scheduled if they have an accumulated delay that is less than the clock period. 

During the execution of the CSTEP algorithm, the primary priority function evaluates data-ready operators for scheduling into the current control step csjj. It first considers the impact of this placement on minimum-time constraints. If scheduling an operator into the current control step violates any minimum-time constraints, this placement is considered too early and must be delayed to a later control step. This is... 

Figure 5-8 shows how this resource utilization checker works in the scheduling process. In the figure, the PLUS operators on the right are data ready for the control step indicated in bold print at the left and have been ordered according to their importance. In the first control step, three operators are data ready. Assume that the initial maxlist value for the PLUS operation is 1, so that at least one plus operator can be scheduled in each control step. The figure shows that operator +a has been scheduled. 

The next row of the figure shows that +b is the next operator to be considered for scheduling. Since the maximum number of prespecified functional units have already been used in the first control step, the resource utilization checker must be invoked. Operator +b is temporarily placed in control step 1, and then the next control step is examined. Control step 2 shows that if +b were to be placed in control step 1, +c which was data ready for control step 1 could be placed in control step 2, and +d, which would be made data ready for control step 2... 

Now +c is considered for placement in control step 1. As shown in the third row of the figure, if +c is placed in control step 1, there are only 2 PLUS operators that can be placed in control step 2. Since a third functional unit cannot be utilized in both control steps 1 and 2, the placement of +c is delayed, even though it is data ready. Since there are no more operators to be considered for placement in control step 1, scheduling proceeds to control step 2. At this point both operators c +d are data ready, and because the maxlist value for PLUS is 2, both can be scheduled without further consideration for resource utilization. Operator + , which is made data ready by operator +c must be delayed until control step 3. The final schedule is listed in row 4 of the figure. 

---